Industrial Talks

[Intel] "Simplifying the Deployment and Verification of HPC Clusters", Ricardo Medel

[Microsoft] "High Performance Computing on the Windows Platform", Markus Christen

[Amd] "Putting all together: the future is fusion", Roberto Brandão

_________________________________

 ABSTRACT

Clusters are the fastest growing segment in HPC. However, clusters remain difficult to specify, deploy, and manage, especially for new HPC users. The Intel Cluster Ready Program goal is to lower the entry barriers to HPC by defining a reference architecture for ensuring hardware/software integration and providing tools that allow end users to get their work done with minimum knowledge about clusters. In this talk we will present the different components of the Intel Cluster Ready Program, with special emphasis on the Intel Cluster Checker tool, a key component of the program.

 SPEAKER BIOGRAPHY

Ricardo Medel received his Bachelor's and Master's degrees in Computer Science from the Universidad Nacional de San Luis (Argentina) and his PhD in Computer Science from the Stevens Institute of Technology (New Jersey, USA). Ricardo was a professor and researcher at the Universidad Nacional de Rio Cuarto for ten years, developing a formal method for verification of graphical designs used in Software Engineering. Since 2006 he works at Intel¿s Argentina Software Development Center, in Cordoba (Argentina). At Intel, he was QA engineer for the XML Security Gateway project and technical leader of the Intel Cluster Checker tool. Currently he is in charge of the Latin America HPC Ecosystem Development area, working with universities, research centers and companies for developing a strong HPC community in Latin America.

_________________________________

 ABSTRACT

In parallel computing scenarios, High Performance Computing (HPC) applications tend to fit within the category of highly distributed scale-out solutions. HPC applications also tend to exhibit tight tolerances for interprocess communication latency and bandwidth metrics. Join this session for a discussion of the "why" and "how" of HPC application development on Windows. Additional this session explores the question of how do we build scalable infrastructure for high performance computing clusters with virtual machines to lower the entry barriers and the complexity to the world of HPC. We discuss who might be interested in such a solution, where it makes sense to build it, pros and cons, how to build it and control it with Microsoft technologies (Windows HPC Server 2008, Hyper-V, Microsoft System Center Virtual Machine Manager and System Center Operations Manager).

 SPEAKER BIOGRAPHY

Markus Christen received his Bachelor's in Computer Science from the University in Zuerich and his post-graduation in Organization from SGO. Markus worked for several years as a Principal Consultant specialist in several HPC implementations around the world and received the Microsoft Circle of Excellence award regarding contributions. Currently he is, working at Microsoft Brazil as an IT Architect, focusing on communities and enterprise customers.

_________________________________

 ABSTRACT

Computer processing has reached a crossroads where the relationship between hardware and software must change to support the increased processing needs of modern computing workloads, including virtualized environments and media-rich applications. Accelerated computing uses specialized hardware to increase the speed of certain processing tasks, offering commercial and consumer users simultaneous energy efficiency, high performance, and low cost. From a hardware perspective, accelerated computing provides a general platform that supports the addition of specialized processing units, or accelerators, to a multicore computer. These accelerators may take the form of off-core chips such as media or network packet accelerators, or they may be additional cores designed for specific types of processing tasks. From a software perspective, the accelerated computing framework will simplify writing parallel applications: developers will have high-level tools that support parallel programming, and applications will not have to specify which hardware should be used for a given processing task.

AMD has been building expertise and capability in general- and specific-purpose computing for the last several years, including how to effectively couple general- and specific-purpose processors. This activity has built a solid foundation for further research and development in accelerated computing. We are actively collaborating with large research institutions, independent software vendors, the open source community, and other organizations to develop an open ecosystem around accelerated computing. Our commitment to accelerated computing includes designing hardware accelerators, building developer tools, and driving thought leadership by participating in standards development bodies. We believe that accelerated computing embodies a rapidly approaching paradigm shift, and we intend to offer commercial and consumer users flexible computing solutions that meet their needs based on the components available in the growing ecosystem.

This presentation explores the accelerated computing framework. It provides examples of accelerated computing in action today, describes potential uses of accelerators in the near- to mid-term, and makes a clear case that accelerated computing is the natural outcome of the computer industry's current integration trend. It further illustrates that, in full bloom, accelerated computing will create many opportunities to customize computer systems by providing targeted, precise solutions for different kinds of workloads to satisfy the need for application performance, energy efficiency, and low cost.